As the CEO of a major telecom network equipment provider, Gary Smith is concerned, along with his customers, about the coming exaflood of Internet traffic, and the ability of today’s networks to scale rapidly and cost-effectively enough to handle it. Smith spoke with Editor-in-Chief Carol Wilson about his optimism that the transition to a software-controlled, highly efficient IP network is well under way.

On growing Internet traffic and greater demand for video quality: You have a multitude of data points around that traffic has started to grow. This is happening as more than 40% of Internet traffic is video and as they put new applications on – and we haven’t started the true mobile Internet yet. I don’t know that it’s doubling every 10 seconds [as some claim] but it’s just common sense that traffic is growing when you look at the multitude of applications. The other issue is quality -- you move from best effort sort of piece where teenaged kids are looking at YouTube to an expectation that is going to increase around quality of that video that you’ve paid for. Consumers are not going to stand for buffering. Plus you have telepresence coming on board as well. The expectations for quality are going to expand.
Even if you say it is going to take longer than you think to develop these applications, there are so many of them.

On the challenge to carriers to scale to meet that capacity need:
I’m optimistic about it. You have developing technologies- there is not magic bullet to any of it, be it DPI [Deep Packet Inspection] or something else out there. They are all part of the armory. There is enough technology innovation to help carriers make this transition. The fundamental challenge is how do you make the transition from voice-architecture networks to IP Ethernet networks that are geared up to carrying this kind of IP and video. I’m reasonably optimistic that technology appears to be coming together to be able to do that and do it economically.

On the cost of that upgrade: We have already seen dramatic improvement in cost per bit bandwidth growth. From 1993 to 2006, we have reduced the cost of long-haul DWDM [dense wave division multiplexing] at a pace that is faster than Moore’s Law. (Smith bases that statement on research done by Dana Cooperson at Ovum-RHK which shows that bandwidth costs per Gigabit per second, per kilometer have dropped from well over $1000 to under $10 as system capacity has gone up.) The overall challenge they’ve got is how do they, even with availability of all of that technology, how to they make the transition, because they have an existing network that is voice-based and terrestrial.